Replies: 3 comments 2 replies
-
|
If i can join in on the question train departing not-quite-understanding. From reading the docs it seems like you actually just… ask Claude to install the packs in order? I did it this way. I cloned the repo, opened Claude, pointed it at the pack.md files in the documented order and in every case, once it read the .md it asked me if I wanted to install it and I said yes, it went off and did the thing. Seemed to run various scripts contained in the pack and do the whole process. It just seemed like an insane way to copy a bunch of files into .claude for me. Took about 20 minutes of letting the model churn in total to install all the packs, and burned quite a bit of tokens doing so. Is there some reason to do it this way? Else why would you not just have an install script? Anyways, it seems like… SOMETHING happened. I actually moved over to Claude code from codex to try out this setup, so I guess I need to swap back to default Claude and see if anything is different. I do notice that the PAI main prompt is loaded as default context, I can see the hooks set up and working in the reasoning trace, the observability dashboard is working, the history/logging system seems to work, and the PAI agent creation skill seems to work also, although on occasion Claude code gets confused and tries to invoke its own agents. I used the personality system to create a separate subagent prompt with the agent creation skill, and that interaction was surprisingly pleasant and useful. I haven’t tested all the features yet, but I guess I’m not certain how to specifically invoke them all or how to verify it is working as intended in every case. It is just bizarre to me to not use an install script, which kind of threw things into doubt if I was being crazy or doing it as intended! Over the past few days, I haven’t had much time to play around with PAI features specifically, but I’ve done a fair bit of my regular coding work in Claude with the PAI system installed but without changing anything about my old work habits, and the Claude code agent seems to operate pretty normally unless I specifically ask to use PAI features. Is this also the intent? |
Beta Was this translation helpful? Give feedback.
-
|
Update: PAI v2.1.0 is now released! 🎉 The installation process has been completely revamped with:
Please try the new installer and let us know if you still experience issues! See: https://github.com/danielmiessler/PAI/releases/tag/v2.1.0 |
Beta Was this translation helpful? Give feedback.
-
|
The v3.0 installer ( |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
I've spent the last couple of days playing around with this and trying to install it. I think I've got most things properly installed.
I've had to make a few modifications with regards to the
GenerateSkillIndex.tsas well as a few other things. I'm still not a hundred percent convinced that it's actually working properly but I also don't have a document to look at to compare and see what I should be expecting to see. I can't seem to get theCOREskill to auto-trigger any of its indexed skills or automatically use stuff like theSkillSearchtool. I installed from scratch on 2.X and CC was not able to accurately follow the instructions and/or there were discrepancies in the instructions such that it confused the AI and not everything got done properly.This is less so an attempt to complain and more so an attempt to understand and get some feedback and guidance on how to get this working and understand what my expectations should be. Would love to hear y'all's thoughts because I'm quite excited to get this going.
Thanks in advance
Beta Was this translation helpful? Give feedback.
All reactions